Zurich Open Repository and Archive University of Zurich Main Library Strickhofstrasse 39 CH-8057 Zurich www.zora.uzh.ch

Year: 2016

A single quantifiable viral load is predictive of virological failure inHuman Immunodeficiency (HIV)-infected patients on combination antiretroviral therapy: The Austrian HIV cohort study

Leierer, Gisela ; Grabmeier-Pfistershammer, Katharina ; Steuer, Andrea ; Sarcletti, Mario ; Geit, Maria ; Haas, Bernhard ; Taylor, Ninon ; Kanatschnig, Manfred ; Rappold, Michaela ; Ledergerber, Bruno ; Zangerle, Robert

Abstract: Background. Viral loads (VLs) detectable at low levels are not uncommon in patients on combination antiretroviral therapy (cART). We investigated whether a single quantifiable VL predicted virological failure (VF). Methods. We analyzed patients receiving standard regimens with at least 1 VL measurement below the limit of quantification (BLQ) in their treatment history. The first VL measure- ment after 6 months of unmodified cART served as baseline VL for the subsequent analyses of thetime to reach single VL levels of ￿200, ￿400, and ￿1000 copies/mL. Roche TaqMan 2.0 was used to quantify human immunodeficiency virus-1 ribonucleic acid. Factors associated with VF were determined byCox proportional hazards models. Results. Of 1614 patients included in the study, 68, 44, and 34 experienced VF ￿200, ￿400, and ￿1000 copies/mL, respectively. In multivariable analyses, compared with patients who were BLQ, a detectable VL ￿ 50 and VL 51-199 copies/mL predicted VF ￿ 200 copies/mL (hazards ratio [HR] = 2.19, 95% confidence interval [CI] = 1.06-4.55 and HR = 4.21, 95% CI = 2.15-8.22, respectively). In those with VL 51-199 copies/mL, a trend for an increased risk of VF ￿400 and VF ￿1000 copies/mL could be found (HR = 2.13, 95% CI = 0.84-5.39 and HR = 2.52, 95% CI = 0.96-6.60, respectively). Con- clusions. These findings support closer monitoring and adherence counseling for patients with asingle measurement of quantifiable VL <200 copies/mL.

DOI: https://doi.org/10.1093/ofid/ofw089

Posted at the Zurich Open Repository and Archive, University of Zurich ZORA URL: https://doi.org/10.5167/uzh-125219 Journal Article Published Version

Originally published at: Leierer, Gisela; Grabmeier-Pfistershammer, Katharina; Steuer, Andrea; Sarcletti, Mario; Geit, Maria; Haas, Bernhard; Taylor, Ninon; Kanatschnig, Manfred; Rappold, Michaela; Ledergerber, Bruno; Zangerle, Robert (2016). A single quantifiable viral load is predictive of virological failure in Human Immunode- ficiency Virus (HIV)-infected patients on combination antiretroviral therapy: The Austrian HIVcohort study. Open Forum Infectious Diseases, 3(2):ofw089. DOI: https://doi.org/10.1093/ofid/ofw089 Open Forum Infectious Diseases MAJOR ARTICLE

A Single Quantifiable Viral Load Is Predictive of Virological Failure in Human Immunodeficiency Virus (HIV)-Infected Patients on Combination Antiretroviral Therapy: The Austrian HIV Cohort Study Gisela Leierer,1,2 Katharina Grabmeier-Pfistershammer,3 Andrea Steuer,4 Mario Sarcletti,1 Maria Geit,5 Bernhard Haas,6 Ninon Taylor,7 Manfred Kanatschnig,8 Michaela Rappold,1,2 Bruno Ledergerber,9 and Robert Zangerle1; for the Austrian HIV Cohort Study Group 1Department of Dermatology and Venereology, Medical University of Innsbruck, 2Austrian HIV Cohort Study, Innsbruck, 3Department of Dermatology, Medical University of Vienna, 4Otto-Wagner Hospital, Vienna, 5Department of Dermatology, General Hospital Linz, 6Department of Internal Medicine, General Hospital Graz-West, 7Department of Internal Medicine III With Hematology, Medical Oncology, Hemostaseology, Infectious Diseases, Rheumatology, Oncologic Center, Laboratory for Immunological and Molecular Cancer Research, Paracelsus Medical University, Salzburg, and 81st Medical Department, General Hospital Klagenfurt, Austria; 9Division of Infectious Diseases and Hospital Epidemiology, University Hospital Zurich, University of Zurich, Switzerland

Background. Viral loads (VLs) detectable at low levels are not uncommon in patients on combination antiretroviral therapy Downloaded from (cART). We investigated whether a single quantifiable VL predicted virological failure (VF). Methods. We analyzed patients receiving standard regimens with at least 1 VL measurement below the limit of quantification (BLQ) in their treatment history. The first VL measurement after 6 months of unmodified cART served as baseline VL for the subse- quent analyses of the time to reach single VL levels of ≥200, ≥400, and ≥1000 copies/mL. Roche TaqMan 2.0 was used to quantify fi

human immunode ciency virus-1 ribonucleic acid. Factors associated with VF were determined by Cox proportional hazards models. http://ofid.oxfordjournals.org/ Results. Of 1614 patients included in the study, 68, 44, and 34 experienced VF ≥200, ≥400, and ≥1000 copies/mL, respectively. In multivariable analyses, compared with patients who were BLQ, a detectable VL ≤ 50 and VL 51–199 copies/mL predicted VF ≥ 200 copies/mL (hazards ratio [HR] = 2.19, 95% confidence interval [CI] = 1.06–4.55 and HR = 4.21, 95% CI = 2.15–8.22, respectively). In those with VL 51–199 copies/mL, a trend for an increased risk of VF ≥400 and VF ≥1000 copies/mL could be found (HR = 2.13, 95% CI = 0.84–5.39 and HR = 2.52, 95% CI = 0.96–6.60, respectively). Conclusions. These findings support closer monitoring and adherence counseling for patients with a single measurement of quan- tifiable VL <200 copies/mL. Keywords.

cART; HIV; low-level viremia; viral load; virological failure. at Zentralbibliothek on August 4, 2016

The goal of combination antiretroviral therapy (cART) is to ob- by a subsequent undetectable VL measurement, or persistent de- tain and maintain viral suppression. Although guidelines have tectable VL [4] or even virological failure (VF). Definitions of VF recommended viral loads (VLs) to be below the limit of quanti- vary between guidelines and, accordingly, some guidelines seem fication (BLQ) of clinically accessible assays [1–4], it remains un- to be more conservative, recommending lower levels of VL (>50 clear whether VLs detectable at low levels, typically below 200 copies/mL) to be reviewed due to a potential risk of rebound [1]. copies/mL, are clinically important with regard to subsequent Others defined VF as having a VL of >200 copies/mL [2– 4], and treatment failures. At the time VL reveals 51 copies/mL or one guideline sets higher thresholds for defining VF, namely VL more, it is uncertain how it will subsequently develop. It may pos- of >1000 copies/mL [5]. Despite several investigations that fo- sibly turn out to be only a viral blip, generally indicating episodes cused on consequences of persistent low-level viremia (LLV), of transiently detectable VL above 50 copies/mL [3, 5] followed which was determined by 2 or more VL measurements [6–12], studies considering 1 single VL measurement to be predictive of VF are very rare. Only 1 study observed patients with a single Received 18 January 2016; accepted 29 April 2016. quantifiable VL below 50 copies/mL to be at higher risk of re- Presented in part: 19th International Workshop on HIV Observational Databases, Catania, Italy; 7th German-Austrian AIDS Conference, Duesseldorf, Germany; 15th European AIDS Con- bound of >50 and >400 copies/mL [13]. Different commercial ference, Barcelona, Spain. techniques are able to detect VLs at low levels. It seems that Correspondence: R. Zangerle, Department of Dermatology and Venereology, Medical Univer- sity of Innsbruck, Anichstrasse 35, 6020 Innsbruck, Austria ([email protected]). low-level, positive VL results appear to be more common with Open Forum Infectious Diseases® some VL assays including the Roche Cobas AmpliPrep/Cobas © The Author 2016. Published by Oxford University Press on behalf of the Infectious Diseases TaqMan 2.0 [14–16]. However, this result needs to be confirmed. Society of America. This is an Open Access article distributed under the terms of the Creative fi Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/licenses/ In this study, we present data from a cohort of well de ned by-nc-nd/4.0/), which permits non-commercial reproduction and distribution of the work, in any human immunodeficiency virus (HIV)-infected patients on un- medium, provided the original work is not altered or transformed in any way, and that the work fi is properly cited. For commercial re-use, please contact [email protected]. modi ed cART with standard regimens over a recent period of DOI: 10.1093/ofid/ofw089 more than 6 months in Austria. The aim of the study was to

Viral Load Predicts Virological Failure • OFID • 1 evaluate the impact of LLV, measured by TaqMan 2.0 assay, and inhibitors (NRTIs) and either a non-NRTI (NNRTI) or a boost- other factors that could increase risk of VF. Virological failure ed protease inhibitor (PI/r) or an inhibitor (INSTI). was defined as HIV ribonucleic acid (RNA) levels of ≥200, Individuals with other cART regimens than those mentioned ≥400, and ≥1000 copies/mL. above were excluded (N = 411). A total of 1010 patients did not have any VL measurement between 6 and 12 months PATIENTS AND METHODS after initiation of the respective cART regimen and were also ex- The Austrian HIV Cohort Study cluded. We further excluded 103 individuals who never had VL The Austrian HIV Cohort Study (AHIVCOS) is an open, multi- BLQ. In the end, 1775 patients fulfilled the inclusion criteria, center, prospective, observational cohort study of HIV-infected and their first VL measurement between 6 and 12 months individuals observed at 7 HIV treatment centers in Austria. after initiation of the respective cART regimen served as base- The study was initiated in 2001 as an incorporated association line for the analyses of the time to single HIV RNA levels of by representatives of 5 Austrian HIV treatments centers (AKH ≥200, ≥400, and ≥1000 copies/mL, respectively (Figure 1). Vienna, Otto-Wagner-Hospital Vienna, AKH Linz, LKH Inns- In order to have a consistent lower limit of quantification of bruck, and LKH Graz West). In 2008, 2 additional HIV treatment 20 copies/mL, 5 of 7 HIV treatment centers that measured VL centers (LKH Salzburg and LKH Klagenfurt) joined AHIVCOS, by the Roche Cobas AmpliPrep/Cobas TaqMan 2.0 (Taqman, thus patients are currently enrolled actively and prospectively at Roche Diagnostics, Mannheim, Germany) were included in

7 public hospital-based HIV treatment centers, which covers the analysis. The remaining 2 centers used the Abbott RealTime Downloaded from approximately 80% of all treated HIV-infected patients in HIV-1 assay (Abbott Diagnostics, Wiesbaden, Germany) and Austria. were excluded. The observation period ended in March 2014. By July 1, 2014, AHIVCOS had included information on Definition of Low-Level Viremia and Virological Failure 8097 individuals. Demographic, clinical, laboratory, and treat- All single quantifiable measurements of <200 copies/mL were http://ofid.oxfordjournals.org/ ment data have been collected prospectively starting at the classified as LLV. For further analyses, LLV was divided into 2 time of inclusion into the cohort. Type of and changes in categories: HIV RNA levels of 51–199 copies/mL and HIV cART, including reasons for interruptions, are recorded, as RNA levels of ≤50 copies/mL. Three definitions for VF were well as the entire medication history including all drugs, dura- used: the “broad” definition reflects a single HIV RNA level tion of intake, and doses patients receive. Acquired immune de- of ≥200 copies/mL, the “restricted” definition includes a single ficiency syndrome events as well as coinfections, such as HIV RNA level of ≥400 copies/mL, and the “stringent” defini- and C, syphilis, etc, were collected. Laboratory pa- tion uses a single HIV RNA level of ≥1000 copies/mL. rameters were recorded continuously and measured in each sin- at Zentralbibliothek on August 4, 2016 gle center. Statistical Methods Baseline data are presented as number (%) or median (inter- Study Design quartile range [IQR]). Characteristics of patients who developed A retrospective analysis was conducted of individuals on cART VF (defined as HIV RNA levels ≥200, ≥400, and ≥1000 copies/ who enrolled in AHIVCOS between July 1, 2012 and July 1, mL) and individuals who did not experience VF in each case 2013. Approval for this study was obtained from the local eth- were compared using χ² tests or Fisher’s exact tests where ap- ical committees of the following participating centers: Vienna propriate for categorical data and non-parametric Wilcoxon Medical University (No. 898/2010), City of Vienna (No. 12- rank-sum tests for continuous data. To assess the predictive 216-VK/2013), Salzburg Federal Government (No. 1159/ value of LLV as well as the influence of various demographic 2010), Graz Medical University (No. 21-431/2010), Innsbruck and clinical parameters on the occurrence of VF, univariable Medical University (No. 283/4.4/2009), Upper Austria Federal and multivariable Cox proportional hazards regression models Government (No. C-3-10/2010), and Carinthian Federal State were applied. For each definition of VF (HIV RNA levels of (No. A-13-11/2011). The patients gave written informed con- ≥200, ≥400, and ≥1000 copies/mL) models were run sepa- sent to store their information in the hospital database and to rately. Multivariable models were adjusted for age and VL at use the data for research. baseline, CD4 count at baseline, HIV transmission category, na- Inclusion Criteria tionality, and prior cART interruptions. Adjusted time-to-event Patients with at least 1 VL measurement BLQ after initiating curves were performed, holding all adjusted covariables fixed at cART had to receive unmodified cART for more than 6 months their mean level. The proportional hazards assumptions were between July 1, 2012 and July 1, 2013 (N = 3299), irrespective of assessed by testing for zero slopes of the scaled Schoenfeld therapies the patients might have taken before this period. Pa- residuals. tients who were not on stable cART due to either interruptions Additional sensitivity analyses were run for different criteria or switches were excluded. During these 6 months, patients in patient selection: one analysis was performed without consid- had to receive cART with 2 nucleoside reverse-transcriptase ering a VL measurement below the limit of detection (BLD) or

2 • OFID • Leierer et al Downloaded from http://ofid.oxfordjournals.org/ at Zentralbibliothek on August 4, 2016

Figure 1. Flowchart of inclusion criteria in patient selection. Abbreviations: BLD, below the limit of detection; BLQ, below the limit of quantification; cART, combination antiretroviral therapy; INSTI, integrase inhibitor; LLV, low-level viremia; NNRTI, nonnucleoside reverse-transcriptase inhibitor; NRTI, nucleoside reverse-transcriptase inhibitor; PI/r, boosted protease inhibitor; VF, virological failure; VL, viral load.

BLQ in treatment history in order to exclude patients with very VL measurement; one analysis was conducted regarding a short cART durations; one analysis was performed excluding shorter recruitment period of 6 months instead of 12 months; patients with cART interruptions or a cART stop after baseline one analysis included patients receiving unmodified cART for

Viral Load Predicts Virological Failure • OFID • 3 >9 months instead of >6 months; and finally, one analysis was In univariable models, several significant factors turned out performed with a shorter recruitment period of 6 months, an to be associated with the development of VF. These factors unmodified cART for >3 months, and without cART interrup- were (1) a higher VL at baseline (compared with patients who tions or a cART stop after baseline VL measurement. All analyses were BLD or BLQ), (2) a lower CD4 cell count at baseline (com- were conducted using Stata software, version 13.1 (StataCorp, pared with patients with ≥500 cells/μL), (3) younger age (<30 College Station, TX). years compared with above 50 years), (4) IDU, (5) women who acquired HIV through heterosexual contact (compared RESULTS with MSM), (6) patient origin from high-prevalence countries

Patient Characteristics Stratified by Virological Failure (compared with low-prevalence countries), (7) prior cART in- A total of 1775 patients fulfilled the inclusion criteria: 79 of terruptions (compared with uninterrupted cART), (8) 2 NRTI them with a VF of ≥200 copies/mL and another 82 of them with a PI/r as cART regimen (compared with 2 NRTI with an fi without any further VL measurement after baseline were ex- NNRTI or INSTI), and nally (9) a higher VL before initiating ≤ cluded. Therefore, 1614 patients were eligible for analyses (Fig- cART (compared with a VL 9.999 copies/mL). ure 1) in this study. Differences in demographic and laboratory Multivariable Cox proportional hazards models for each def- characteristics between the patients who developed VF and in- inition of VF were calculated and controlled for VL at baseline, dividuals who did not experience VF are described in Table 1. CD4 count at baseline, age at baseline, HIV transmission cate- Stratification was made according to the 3 definitions of VF; gory, nationality, and prior cART interruptions. The resulting Downloaded from therefore, 68 (4.2%), 44 (2.7%), and 34 (2.1%) of 1614 incorpo- adjusted cumulative hazard curves are shown in Figure 2. Com- rated patients progressed to a VF defined as an HIV RNA level pared with patients who were BLD or BLQ, patients with a fi ≤ of ≥200, ≥400, and ≥1000 copies/mL, respectively. Median fol- quanti able VL of 50 copies/mL at baseline had a 2.19 times ≥ higher risk of VF of ≥200 copies/mL; however, this did not en- low-up time for all patients regarding VF of 200 copies/mL as http://ofid.oxfordjournals.org/ fi endpoint was 9.2 months (IQR, 7.8–11.5). Each VF group dif- compass the restricted and stringent de nition of VF. Viral load fered significantly from the non-VF group in VL at baseline, age between 51 and 199 copies/mL at baseline also increased the ≥ at baseline, HIV transmission category, CD4 count at baseline, risk of VF 200 copies/mL, compared with patients who were – prior cART interruptions, as well as cART regimen. Compared BLD or BLQ at baseline (HR = 4.21, 95% CI = 2.15 8.22). In ad- ≥ ≥ with individuals who experienced VF, patients who did not de- dition, a trend for a higher risk of VF 400 and 1000 copies/ – velop a VF were older and had a higher CD4 cell count at base- mL could be found (HR = 2.13, 95% CI = 0.84 5.39 and – line. The proportion of injecting drug users (IDUs) was higher, HR = 2.52, 95% CI = 0.96 6.60, respectively). The association and there were fewer men who have sex with men (MSM) in the between lower CD4 count at baseline and the development of at Zentralbibliothek on August 4, 2016 VF group. Patients who had a VF ≥ 400 and ≥1000 copies/mL VF found in univariable analyses vanished after adjustment. originated more frequently from high-prevalence countries. Pa- The hazard ratio for VF was higher in patients under 30 years tients with VF had more prior cART interruptions, were more compared with individuals older than 50 years: HIV RNA ≥ – often on 2 NRTIs with a PI/r instead of a NNRTI or an INSTI, levels of 200 copies/mL, HR = 2.98 and 95% CI = 1.24 7.21; ≥ and VL at baseline was higher compared with individuals who HIV RNA levels of 400 copies/mL, HR = 3.67 and 95% – ≥ did not experience VF. More importantly, in patients who had a CI = 1.13 11.96; HIV RNA levels of 1000 copies/mL, – VF defined as HIV RNA levels of ≥200 copies/mL, VL before HR = 5.02 and 95% CI = .20 21.04, respectively. Injecting initiating cART was higher compared with the non-VF group, drug use was associated with an increased risk of VF in ≥ but not for patients with VF defined as HIV RNA level of ≥400 males: HIV RNA levels of 200 copies/mL, HR = 3.21 and – ≥ and ≥1000 copies/mL. No difference between the groups could 95% CI = 1.50 6.88; HIV RNA levels of 400 copies/mL, – be found in ever having had a VF before VL at baseline, in prior HR = 4.57 and 95% CI = 1.85 11.33; HIV RNA levels of ≥ – cART duration and in first-line cART of the respective regimen. 1000 copies/mL, HR = 5.25 and 95% CI = 1.81 15.22, respec- tively, compared with MSM. Injecting drug use was also associ- Predictors of Virological Failure ated with an increased risk of VF in females: HIV RNA levels of Table 2 provides results from univariable and multivariable Cox ≥200 copies/mL, HR = 3.83 and 95% CI = 1.25–11.74; HIV proportional hazards models for the occurrence of VF accord- RNA levels of ≥400 copies/mL, HR = 4.52 and 95% CI = 1.20– ing to the broad definition of VF (defined as HIV RNA level of 17.03, respectively, compared with MSM. Female heterosexuals ≥200 copies/mL). Hazard ratios (HRs) for the development of showed a higher risk of VF of ≥200 copies/mL (HR = 2.25, VF regarding the restricted and stringent definition of VF (de- 95% CI = 1.08–4.66), but not for VF defined by higher VL levels. fined as HIV RNA level of ≥400 and ≥1000 copies/mL, respec- Patients originating from high-prevalence countries were at tively) are shown in the Supplemental Table 1. A total of 1614 higher risk to develop VF ≥1000 copies/mL compared with patients were observed until the development of VF or last VL individuals from low-prevalence countries (HR = 2.94, 95% measurement or death, whichever came first. CI = 1.03–8.43). The hazard ratio of VF for all 3 definitions was

4 • OFID • Leierer et al Table 1. Characteristics of Patients Stratified by Virological Failure Defined as HIV RNA Levels ≥200, ≥400, and ≥1000 copies/mL vs Nonvirological Failure in Each Casea

All Patients (N = 1614) All Patients (N = 1614) All Patients (N = 1614)

Broad Definition Restricted Definition Stringent Definition

VL < 200 VF ≥ 200 VL < 400 VF ≥ 400 VL < 1000 VF ≥ 1000

N = 1546 N = 68 N = 1570 N = 44 N = 1580 N = 34

No. of Patients N (%) N (%) P Value N (%) N (%) P Value N (%) N (%) P Value

VL at baseline <.001 .012 .008 51–199 copies/mL 62 (4.0) 13 (19.1) 69 (4.4) 6 (13.6) 69 (4.4) 6 (17.7) ≤50 copies/mL 103 (6.7) 9 (13.2) 108 (6.9) 4 (9.1) 110 (7.0) 2 (5.9) BLD + BLQ 1381 (89.3) 46 (67.7) 1393 (88.7) 34 (77.3) 1401 (88.7) 26 (76.5) Age at baseline .028 .045 .019 <30 yr 141 (9.1) 11 (16.2) 145 (9.3) 7 (15.9) 146 (9.3) 6 (17.7) 30–50 yr 996 (64.4) 47 (69.1) 1011 (64.4) 32 (72.7) 1018 (64.4) 25 (73.5) >50 yr 409 (26.5) 10 (14.7) 414 (26.4) 5 (11.4) 416 (26.3) 3 (8.8) HIV transmission category <.001 <.001 .001 Male injecting drug user 133 (8.6) 14 (20.6) 135 (8.6) 12 (27.3) 137 (8.7) 10 (29.4)

Female injecting drug user 46 (3.0) 4 (5.9) 47 (3.0) 3 (6.8) 48 (3.0) 2 (5.9) Downloaded from Male heterosexual 340 (22.0) 8 (11.8) 343 (21.9) 5 (11.4) 343 (21.7) 5 (14.7) Female heterosexual 339 (21.9) 21 (30.9) 348 (22.2) 12 (27.3) 351 (22.2) 9 (26.5) Other 65 (4.2) 6 (8.8) 68 (4.3) 3 (6.8) 69 (4.4) 2 (5.9) Men who have sex with men 623 (40.3) 15 (22.1) 629 (40.1) 9 (20.5) 632 (40.0) 6 (17.7) Nationality .133 .012 .005 High prevalence country 143 (9.3) 10 (14.7) 144 (9.2) 9 (20.5) 145 (9.2) 8 (23.5) http://ofid.oxfordjournals.org/ Low prevalence country 1403 (90.8) 58 (85.3) 1426 (90.8) 35 (79.6) 1435 (90.8) 26 (76.5) CD4 count at baseline .001 .015 .015 Missing 41 (2.7) 7 (10.3) 44 (2.8) 4 (9.1) 45 (2.9) 3 (8.8) <200 cells/µL 48 (3.1) 5 (7.4) 49 (3.1) 4 (9.1) 49 (3.1) 4 (11.8) 200–349 cells/µL 174 (11.3) 7 (10.3) 176 (11.2) 5 (11.4) 176 (11.1) 5 (14.7) 350–499 cells/µL 264 (17.1) 12 (17.7) 267 (17.0) 9 (20.5) 271 (17.2) 5 (14.7) ≥500 cells/µL 1019 (65.9) 37 (54.4) 1034 (65.9) 22 (50.0) 1039 (65.8) 17 (50.0) Prior cART interruptionsb <.001 <.001 <.001

≥1 339 (21.9) 30 (44.1) 348 (22.2) 21 (47.7) 352 (22.3) 17 (50.0) at Zentralbibliothek on August 4, 2016 None 1207 (78.1) 38 (55.9) 1222 (77.8) 23 (52.3) 1228 (77.7) 17 (50.0) cART regimen .001 .024 .026 2 NRTI + PI/r 611 (39.5) 40 (58.8) 626 (39.9) 25 (56.8) 631 (39.9) 20 (58.8) 2 NRTI + NNRTI/INSTI 935 (60.5) 28 (41.2) 944 (60.1) 19 (43.2) 949 (60.1) 14 (41.2) Ever VF before VL at baseline .140 .340 .373 <200 copies/mL 83 (5.4) 2 (2.9) 83 (5.3) 2 (4.6) 84 (5.3) 1 (2.9) 200–399 copies/mL 416 (26.9) 26 (38.2) 428 (27.3) 14 (31.8) 430 (27.2) 12 (35.3) 400–999 copies/mL 303 (19.6) 8 (11.8) 307 (19.6) 4 (9.1) 308 (19.5) 3 (8.8) ≥1000 copies/mL 744 (48.1) 32 (47.1) 752 (47.9) 24 (54.6) 758 (48.0) 18 (52.9) VL before initiating cART .045 .399 .330 Missing 272 (17.6) 13 (19.1) 277 (17.6) 8 (18.2) 280 (17.7) 5 (14.7) >99.999 copies/mL 559 (36.2) 34 (50.0) 573 (36.5) 20 (45.5) 577 (36.5) 16 (47.1) 10.000–99.999 copies/mL 527 (34.1) 18 (26.5) 531 (33.8) 14 (31.8) 533 (33.7) 12 (35.3) ≤9.999 copies/mL 188 (12.2) 3 (4.4) 189 (12.0) 2 (4.6) 190 (12.0) 1 (2.9) Prior cART durationc 1.000 .873 .710 <9 mo 61 (4.0) 2 (2.9) 61 (3.9) 2 (4.6) 61 (3.9) 2 (5.9) 9–18 mo 139 (9.0) 6 (8.8) 141 (9.0) 4 (9.1) 142 (9.0) 3 (8.8) >18 mo 1346 (87.1) 60 (88.2) 1368 (87.1) 38 (86.4) 1377 (87.2) 29 (85.3) Antiretroviral therapy-naived .146 .218 .181 Yes 290 (18.8) 8 (11.8) 293 (18.7) 5 (11.4) 295 (18.7) 3 (8.8) No 1256 (81.2) 60 (88.2) 1277 (81.3) 39 (88.6) 1285 (81.3) 31 (91.2) Follow-up time in months 9.3 (8.0–11.5) 6.4 (3.3–8.2) <.001 9.4 (8.0–11.5) 7.0 (5.0–8.4) <.001 9.4 (8.0–11.6) 7.1 (5.8–8.7) <.001

Abbreviations: BLD, below the limit of detection; BLQ, below the limit of quantification; cART, combination antiretroviral therapy; HIV, human immunodeficiency virus; INSTI, integrase inhibitor; NNRTI, nonnucleoside reverse-transcriptase inhibitor; NRTI, nucleoside reverse-transcriptase inhibitor; PI/r, boosted protease inhibitor; RNA, ribonucleic acid; VF, virological failure; VL, viral load. a The first VL measurement after baseline VL measurement (baseline: the first detectable VL measurement within 6 months after receiving unmodified cART for >6 months with standard regimens between July 2012 and 2013, excluding patients without any VL measurement BLQ in their treatment history as well as patients without any baseline VL measurement). [Number (%) or median (interquartile range)]. b Interruptions prior to 6 months of unmodified cART of the respective cART regimen. c cART duration until 6 months of unmodified cART of the respective cART regimen. d Whether the respective cART regimen is a first-line cART or not.

Viral Load Predicts Virological Failure • OFID • 5 Table 2. Univariable and Multivariable Cox Regression Results: Association Between Different Factors and Virological Failure Defined as HIV RNA Levels ≥200 Copies/mL (Broad Definition)

Outcome VF ≥ 200

No. of Patients Included N = 1614

No. of Failures N = 68

Univariable Multivariable

HR (95% CI) HR (95% CI)

VL at baseline 51–199 copies/mL 5.84 (3.15–10.82) 4.21 (2.15–8.22) ≤50 copies/mL 2.47 (1.21–5.05) 2.19 (1.06–4.55) BLD + BLQ 1.00 (Reference) 1.00 (Reference) Age at baseline <30 y 3.31 (1.41–7.81) 2.98 (1.24–7.21) 30–50 yr 1.94 (.98–3.84) 1.91 (.95–3.84) >50 yr 1.00 (Reference) 1.00 (Reference) HIV transmission category

Male injecting drug user 4.53 (2.19–9.40) 3.21 (1.50–6.88) Downloaded from Female injecting drug user 3.83 (1.27–11.54) 3.83 (1.25–11.74) Male heterosexual 1.00 (.42–2.35) 1.07 (.44–2.59) Female heterosexual 2.64 (1.36–5.13) 2.25 (1.08–4.66) Other 3.72 (1.44–9.59) 3.66 (1.40–9.56) Men who have sex with men 1.00 (Reference) 1.00 (Reference) Nationality http://ofid.oxfordjournals.org/ High prevalence country 1.62 (.83–3.16) 1.07 (.48–2.36) Low prevalence country 1.00 (Reference) 1.00 (Reference) CD4 count at baseline Missing 4.60 (2.05–10.32) 3.31 (1.40–7.81) <200 cells/µL 2.91 (1.14–7.41) 1.66 (.62–4.43) 200–349 cells/µL 1.11 (.50–2.49) 0.95 (.42–2.16) 350–499 cells/µL 1.25 (.65–2.39) 1.48 (.76–2.87) ≥500 cells/µL 1.00 (Reference) 1.00 (Reference) a Prior cART interruptions at Zentralbibliothek on August 4, 2016 ≥1 2.80 (1.73–4.52) 2.28 (1.37–3.77) None 1.00 (Reference) 1.00 (Reference) cART regimen 2 NRTI + PI/r 2.13 (1.31–3.45) 2 NRTI + NNRTI/INSTI 1.00 (Reference) Ever VF before VL at baseline <200 copies/mL 0.58 (.14–2.42) 200–399 copies/mL 1.40 (.84–2.36) 400–999 copies/mL 0.59 (.27–1.29) ≥1000 copies/mL 1.00 (Reference) VL before initiating cART Missing 2.91 (.83–10.21) >99.999 copies/mL 3.49 (1.07–11.37) 10.000–99.999 copies/mL 2.10 (.62–7.14) ≤9.999 copies/mL 1.00 (Reference) Prior cART durationb <9 mo 0.85 (.21–3.47) Figure 2. Adjusted cumulative hazard curves for the development of virological 9–18 mo 0.96 (.42–2.22) failure based on a single viral load at baseline. Abbreviations: BLD, below the >18 mo 1.00 (Reference) limit of detection; BLQ, below the limit of quantification; CI, confidence interval; Antiretroviral therapy-naivec HR, hazards ratio; LLV, low-level viremia. Yes 0.55 (.26–1.15) No 1.00 (Reference)

Abbreviations: BLD, below the limit of detection; BLQ, below the limit of quantification; cART, combination antiretroviral therapy; CI, confidence interval; HIV, human immunodeficiency higher in patients who have had prior cART interruptions: HIV virus; HR, hazard ratio; INSTI, integrase inhibitor; NNRTI, nonnucleoside reverse- RNA levels of ≥200 copies/mL, HR = 2.28 and 95% CI = 1.37– transcriptase inhibitor; NRTI, nucleoside reverse-transcriptase inhibitor; PI/r, boosted protease inhibitor; RNA, ribonucleic acid; VF, virological failure; VL, viral load. 3.77; HIV RNA levels ≥400 copies/mL, HR = 2.77 and 95% a Interruptions prior to 6 months of unmodified cART of the respective cART regimen. CI = 1.48–5.19; HIV RNA levels ≥1000 copies/mL, HR = 2.88 b cART duration until 6 months of unmodified cART of the respective cART regimen. c Whether the respective cART regimen is a first-line cART or not. and 95% CI = 1.41–5.87. However, all additional analyses did

6 • OFID • Leierer et al not reveal any substantial differences in HRs compared with our adherence. Treatment interruptions, defined as cART discon- primary analysis (data not shown). tinuation for at least 8 days after having started cART, were ob- served more frequently in patients who developed VF DISCUSSION (approximately 44% vs 22%). In adjusted analyses, the risk of In this study, we specifically investigated the impact of LLV on VF of ≥200 was 2.3-fold higher in patients who had prior VF, defined as HIV RNA level of ≥200, ≥400, and ≥1000 cop- cART interruptions compared with those who never interrupt- ies/mL, among a cohort of HIV-positive individuals in Austria ed treatment. As a consequence, we demonstrated reduced ad- on unmodified cART with standard regimens over a recent pe- herenceasapredictorofVF.Doyleetal[13]foundmajor riod of more than 6 months. Other potential risk factors were resistance-associated mutations in less than half of the patients also studied. with rebound of >400 copies/mL, and they concluded that We observed that 1 single measurement of quantifiable VL poor adherence was as a cause of rebound. A decline in VF below 200 copies/mL under unmodified cART predicted the oc- was shown for patients who have had virologic suppression for currence of VF, according to the broad definition of VF (≥200 longer time periods than individuals suppressed for shorter peri- copies/mL). The risk to develop such VF was 2.19 times higher ods [19, 20]. However, in our study, prior cART duration did not for individuals whose VL was ≤50 copies/mL and 4.21 times predict VF. We recently found that duration of cART is signifi- higher for those with VL of 51–199 copies/mL compared with cantly associated with LLV [21], which is in line with other re- patients who were BLD or BLQ at baseline. This finding was in- ports [22, 23] but not all [24]. Downloaded from dependent of other covariates including CD4 count at baseline, Furthermore, we found that the type of cART regimen was age at baseline, HIV transmission category, nationality, and associated with low-level rebound [8] and that NNRTI-based prior cART interruptions, which may influence the outcome. regimens were able to exert a stronger inhibitory effect on Regarding the restricted and the stringent definition of VF, a than PI-based regimens [13, 25, 26]. In a previ- http://ofid.oxfordjournals.org/ trend for an increased risk of VF could be found in patients hav- ous study, we showed that patients BLD or BLQ were more like- ing had a VL of 51–199 copies/mL at baseline. ly to be treated with 2 NRTIs and an NNRTI or INSTI Our results extend and refine data from several recent studies compared with those with LLV, most of whom tended to be showing the predictive value of LLV. Although these studies dif- on 2 NRTIs and a PI/r. In contrast, we found a 1.5-fold fer in study designs and methods, they all conclude that persis- increased risk of LLV for patients receiving cART with a PI/ tent LLV (determined by repeated measurements) elevates the r-based regimen compared with those on a NNRTI- or INSTI- risk of VF [6–11]. based regimen [21]. However, this effect could be confounded

In clinical routine healthcare, providers encounter 1 single by a selection bias because PI/r-based cART regimens may pref- at Zentralbibliothek on August 4, 2016 measurement of VL. Only 1 study has hitherto addressed this erentially be prescribed to individuals with concerns regarding important issue; however, a different commercial HIV RNA adherence. A recent retrospective study in patients on cART assay was used in this study, namely the Abbott RealTime with viremia between 50 and 100 copies/mL found that plasma assay. It pointed out the strong predictive value for VF in pa- drug levels at the first LLV episode were associated with subse- tients with 1 single VL measurement below 50 copies/mL quent VF, which was defined as HIV RNA ≥1000 copies/mL. In [13]. Our study, based on the Roche Cobas AmpliPrep/Cobas addition, it was shown that together with resistance data, a high- TaqMan 2.0 assay, essentially confirms these data, but we pro- er proportion of treatment failures can be explained than either vide more detail and use this in conjunction with other param- measure alone [27]. However, at this time, HIV cannot be erad- eters. In our study, the risk of the development of VF decreased icated by cART; therefore, despite being on stable cART, it re- with increasing age and escalated with stricter threshold defini- mains a challenge for some patients to achieve HIV RNA levels tions of VF. However, our finding that younger age was an in- BLD or BLQ [28]. Low-level viremia might arise from the fol- dependent factor of viral rebound in patients with a single lowing: (1) ongoing cycles of viral replication in a sanctuary site measurement of VL <50 copies/mL was also confirmed by where drug levels are suboptimal; (2) long-lived infected cells other studies [17, 18]; however, these were studies on persistent that produce virus; or (3) activation of virus expression from la- LLV. In addition, we demonstrated a higher risk of VF in male tently infected CD4 T-cell reservoirs [29]. Studies on the effect of and female IDUs compared with MSM. The risk was higher in cART intensification in patients with LLV did not find evidence men and increased with higher VF thresholds. In our study, pa- of ongoing replication [30], whereas others concluded that active tients from high-prevalence countries had an almost 3-fold higher replication persists in some individuals on cART [31]. In our risk of experiencing VF ≥1000 copies/mL after a single measure- study, only a small proportion of the patients with VL BLD or ment of LLV, similar to the correlation to persistent LLV [17]. BLQ at baseline experienced a VF of ≥200 copies/mL (3%). Because adherence to therapy is difficult to measure accurate- A major strength of this study is the open, observational design ly, cART interruptions before 6 months under stable cART of with complete follow up. Our study provides data of HIV-positive the respective cART regimen were considered as a proxy for individuals in Austria and is representative of an entire country;

Viral Load Predicts Virological Failure • OFID • 7 therefore, selection bias was minimized. This study has several lim- 4. Gunthard HF, Aberg JA, Eron JJ, et al. Antiretroviral treatment of adult HIV in- fection: 2014 recommendations of the International Antiviral Society-USA Panel. itations. There was a small number of VFs, especially with regard JAMA 2014; 312:410–25. to the stringent definition, which limits the number of variables for 5. Consolidated Guidelines on the Use of Antiretroviral Drugs for Treating ’ which the analysis could be adjusted. We were not able to adjust and Preventing HIV Infection: What s New. World Health Organization. Geneva; 2015. Available at: http://apps.who.int/iris/bitstream/10665/198064/1/ for several parameters that might influence cART pharmacokinet- 9789241509893_eng.pdf?ua=1. Accessed 25 March 2016. icssuchasbodymassindexandrenalfunction.However,noas- 6. Antiretroviral Therapy Cohort Collaboration (ART-CC), Vandenhende MA, Ingle S, et al. Impact of low-level viremia on clinical and virological outcomes in treated sociation could be found between these factors and LLV [7]. HIV-1-infected patients. AIDS 2015; 29:373–83. Furthermore, we were not able to adjust for socioeconomic and 7. Boillat-Blanco N, Darling KE, Schoni-Affolter F, et al. Virological fl outcome and management of persistent low-level viraemia in HIV-1-infected lifestyle factors such as smoking, which may also in uence the re- patients: 11 years of the Swiss HIV Cohort Study. Antivir Ther 2015;20: sults, because of missing or incomplete data. Moreover, we did not 165–75. 8. Geretti AM, Smith C, Haberl A, et al. Determinants of virological failure after suc- analyze the association between VF and the development of resis- cessful viral load suppression in first-line highly active antiretroviral therapy. Anti- tance mutations. Previous studies documented the risk of acquir- vir Ther 2008; 13:927–36. ing resistance mutations during LLV [32, 33]andinpatientswho 9. Laprise C, de Pokomandy A, Baril JG, et al. Virologic failure following persistent low-level viremia in a cohort of HIV-positive patients: results from 12 years of ob- experienced rebound [13]. servation. Clin Infect Dis 2013; 57:1489–96. 10. Pernas B, Grandal M, Pertega S, et al. Any impact of blips and low-level viraemia CONCLUSIONS episodes among HIV-infected patients with sustained virological suppression on ART? J Antimicrob Chemother 2016; 71:1051–5.

This study of well defined patients on unmodified cART over a 11. Sungkanuparph S, Groger RK, Overton ET, et al. Persistent low-level viraemia and Downloaded from period of more than 6 months gives insights into different pre- virological failure in HIV-1-infected patients treated with highly active antiretro- viral therapy. HIV Med 2006; 7:437–41. dictors for the development of VF. Our main finding of an in- 12. Vandenhende MA, Perrier A, Bonnet F, et al. Risk of virological failure in HIV-1- creased risk of VF in patients with 1 single quantifiable VL infected patients experiencing low-level viraemia under active antiretroviral ther- apy (ANRS C03 cohort study). Antivir Ther 2015; 20:655–60. measurement below 200 copies/mL supports the need to evaluate 13. Doyle T, Smith C, Vitiello P, et al. Plasma HIV-1 RNA detection below 50 copies/ strategies for managing LLV in cART-treated patients. These re- ml and risk of virologic rebound in patients receiving highly active antiretroviral http://ofid.oxfordjournals.org/ therapy. Clin Infect Dis 2012; 54:724–32. sults provide implications for patient management by emphasiz- 14. Sire JM, Vray M, Merzouk M, et al. Comparative RNA quantification of HIV-1 ing closer monitoring and adherence counseling. Furthermore, group M and non-M with the Roche Cobas AmpliPrep/Cobas TaqMan HIV-1 fi v2.0 and Abbott Real-Time HIV-1 PCR assays. J Acquir Immune DeficSyndr our ndings could form the basis for future methods and strat- 2011; 56:239–43. egies to improve the outcome in single quantifiable VL studies. 15. Taylor N, Grabmeier-Pfistershammer K, Egle A, et al. Cobas ampliprep/cobas Taq- Man HIV-1 v2.0 assay: consequences at the cohort level. PLoS One 2013; 8:e74024. Supplementary Data 16. Swenson LC, Cobb B, Geretti AM, et al. Comparative performances of HIV-1 RNA load assays at low viral load levels: results of an international collaboration. J Clin Supplementary material is available online at Open Forum Infectious Diseases Microbiol 2014; 52:517–23. at Zentralbibliothek on August 4, 2016 online (http://OpenForumInfectiousDiseases.oxfordjournals.org/). 17. Bansi LK, Benzie AA, Phillips AN, et al. Are previous treatment interruptions as- sociated with higher viral rebound rates in patients with viral suppression? AIDS Acknowledgments 2008; 22:349–56. We are grateful to all clinicians, data managers, and research nurses in 18. Henrich TJ, Wood BR, Kuritzkes DR. Increased risk of virologic rebound in pa- participating human immunodeficiency virus (HIV) treatments centers listed tients on antiviral therapy with a detectable HIV load <48 copies/mL. PLoS One in the Appendix. Furthermore, a special thank goes to DI Heinz Appoyer 2012; 7:e50065. (now called network vita) who developed the HIV Patient Management 19. Lima VD, Bangsberg DR, Harrigan PR, et al. Risk of viral failure declines with du- ration of suppression on highly active antiretroviral therapy irrespective of adher- System. ence level. J Acquir Immune Defic Syndr 2010; 55:460–5. Disclaimer. Companies involved in marketing HIV drugs (AbbVie, 20. Rosenblum M, Deeks SG, van der Laan M, Bangsberg DR. The risk of virologic Boehringer-Ingelheim, Bristol-Myers-Squibb, GILEAD, GSK Austria on failure decreases with duration of HIV suppression, at greater than 50% adherence behalf of Viif, Janssen, MSD) provided equal contributions, irrespective of to antiretroviral therapy. PLoS One 2009; 4:e7196. their market shares. 21. Leierer G, Grabmeier-Pfistershammer K, Steuer A, et al. Factors associated with Financial support. This work was funded by the Austrian Agency for low-level viraemia and virological failure: results from the Austrian HIV cohort Health and Food Safety (AGES), participating hospitals running HIV treat- study. PLoS One 2015; 10:e0142923. ment centers, and partners in the pharmaceutical industry. 22. Pascual-Pareja JF, Martinez-Prats L, Luczkowiak J, et al. Detection of HIV-1 at be- Potential conflicts of interest. All authors: No reported conflicts. All tween 20 and 49 copies per milliliter by the Cobas TaqMan HIV-1 v2.0 assay is associated with higher pretherapy viral load and less time on antiretroviral therapy. authors have submitted the ICMJE Form for Disclosure of Potential Con- – fl J Clin Microbiol 2010; 48:1911 2. icts of Interest. 23. Vancoillie L, Demecheleer E, Callens S, et al. Markers associated with persisting low-level viraemia under antiretroviral therapy in HIV-1 infection. J Antimicrob References Chemother 2014; 69:1098–103. 1. European Guidelines for treatment of HIV-infected adults in Europe, Version 8.0, 24. Charpentier C, Landman R, Laouenan C, et al. Persistent low-level HIV-1 RNA October 2015. European AIDS Clinical Society. Brussels, Belgium, 2015. Available between 20 and 50 copies/mL in antiretroviral-treated patients: associated factors at: http://www.eacsociety.org/files/2015_eacsguidelines_8.0-english_revised- and virological outcome. J Antimicrob Chemother 2012; 67:2231–5. 20151104.pdf. Accessed 25 March 2016. 25. Bonora S, Nicastri E, Calcagno A, et al. Ultrasensitive assessment of residual HIV 2. British HIV Association guidelines for the treatment of HIV-1-positive adults viraemia in HAART-treated patients with persistently undetectable plasma HIV- with antiretroviral therapy 2015. British HIV Association, 2015. Available at: RNA: a cross-sectional evaluation. J Med Virol 2009; 81:400–5. http://www.bhiva.org/documents/Guidelines/Treatment/2015/2015-treatment- 26. Widdrington J, Payne B, Medhi M, et al. The significance of very low-level virae- guidelines.pdf. Accessed 25 March 2016. mia detected by sensitive viral load assays in HIV infected patients on HAART. 3. Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the J Infect 2011; 62:87–92. use of antiretroviral agents in HIV-1-infected adults and adolescents. Department 27. Gonzalez-Serna A. Untimed drug levels and resistance in patients experiencing of Health and Human Services. Rockville, MD, 2016. Available at: https://aidsinfo. low-level viremia. In: Conference on and Opportunistic Infections. nih.gov/contentfiles/lvguidelines/adultandadolescentgl.pdf. Accessed 25 March 2016. Seattle, Washington. February 23–26, 2015. (Abstract 117).

8 • OFID • Leierer et al 28. Palmer S, Maldarelli F, Wiegand A, et al. Low-level viremia persists for at least HIV Treatment Centers: LKH Innsbruck (coordinating cen- 7 years in patients on suppressive antiretroviral therapy. Proc Natl Acad Sci U S A 2008; 105:3879–84. ter): Martin Gisinger, Maria Kitchen, Elisabeth Rieser, Brigitte 29. Palmer S, Josefsson L, Coffin JM. HIV reservoirs and the possibility of a cure for Rühr, Mario Sarcletti, Robert Zangerle. LKH Salzburg: Alexander – HIV infection. J Intern Med 2011; 270:550 60. Egle, Richard Greil, Michaela Schachner, Ninon Taylor. AKH 30. McMahon D, Jones J, Wiegand A, et al. Short-course raltegravir intensification does not reduce persistent low-level viremia in patients with HIV-1 suppression Linz: Jörg Berg, Maria Geit, Angela Öllinger. AKH Vienna: Regi- during receipt of combination antiretroviral therapy. Clin Infect Dis 2010; na Aichwalder, Katharina Grabmeier-Pfistershammer, Armin 50:912–9. 31. Buzon MJ, Massanella M, Llibre JM, et al. HIV-1 replication and immune dynam- Rieger, Veronique Touzeau. Otto-Wagner Hospital Vienna: ics are affected by raltegravir intensification of HAART-suppressed subjects. Nat Piotr Cichon, Manfred Gartner, Brigitte Schmied, Andrea Steuer. Med 2010; 16:460–5. 32. Nettles RE, Kieffer TL, Simmons RP, et al. Genotypic resistance in HIV-1-infected LKH Graz-West: Bernhard Haas, Andreas Kapper, Elmar Wall- patients with persistently detectable low-level viremia while receiving highly active ner. LKH Klagenfurt: Silvana Achatz, Manfred Kanatschnig, antiretroviral therapy. Clin Infect Dis 2004; 39:1030–7. 33. Taiwo B, Gallien S, Aga E, et al. Antiretroviral drug resistance in HIV-1-infected Georg Schober. patients experiencing persistent low-level viremia during first-line therapy. J Infect : Elisabeth Puchhammer-Stöckl (Vienna). – Dis 2011; 204:515 20. Data Management Group: Heinz Appoyer (IT-related), Gisela Leierer (AHIVCOS), Michaela Rappold (AHIVCOS), APPENDIX Stefanie Strickner (AHIVCOS). The Austrian HIV Cohort Study ([AHIVCOS] Steering Com- Data Safety and Protection: Klaus Schindelwig (Innsbruck). mittee): Alexander Egle, Maria Geit, Bernhard Haas, Manfred Ka- Scientific Advisory Board: Bruno Ledergerber (Zurich), Downloaded from natschnig, Armin Rieger, Andrea Steuer, Robert Zangerle (Chair). Gerd Fätkenheuer (Cologne). http://ofid.oxfordjournals.org/ at Zentralbibliothek on August 4, 2016

Viral Load Predicts Virological Failure • OFID • 9